Monitoring the online customer experience, by Mark Hurst.
More Resources

Monthly updates via e-mail: to sign up.

Download Managing Incoming E-mail report

Interview: Marissa Mayer, Product Manager, Google

Tuesday, October 15, 2002
by Mark Hurst
(esta página en español)

Note: Marissa Mayer will speak at the Gel (Good Experience Live) conference, May 2003, in New York City. Details here. has been a leader in online experience for some time. I recently conducted a phone interview with Marissa Mayer, Google product manager, to learn how Google creates and improves its customer experience.

Marissa Mayer was the first person at Google to work exclusively on the user experience, starting in 2000. The UI team now consists of about eight people.

Q: What is Google's secret in maintaining such a simple and clear user experience, so consistently?

Larry Page, our founder, has been very important because of his background in HCI. We don't have to play the politics, like I've heard you have to do in other companies. Google executives are very understanding of user experience needs. They want to understand what happened on a detailed level. We'll tell them, we had eight users come in for a user test, and five couldn't use this feature.

All of us on the UI team think the value of Google is in not being cluttered, in offering a great user experience. I like to say that Google should be "what you want, when you want it." As opposed to "everything you could ever want, even when you don't."

I think Google should be like a Swiss Army knife: clean, simple, the tool you want to take everywhere. When you need a certain tool, you can pull these lovely doodads out of it and get what you want. So on Google, rather than showing you upfront that we can do all these things, we give you tips to encourage you to do things these ways. We get you to put your query in the search field, rather than have all these links up front. That's worked well for us. Like when you see a knife with all 681 functions opened up, you're terrified. That's how other sites are - you're scared to use them. Google has that same level of complexity, but we have a simple and functional interface on it, like the Swiss Army knife closed.

The utmost thing is the user experience, to have the most useful experience. It's important to differentiate between "usefulness" and "usability." At Google, we make a *useful* tool, and then we put a *usable* interface on top of that. One has to precede the other. If you have usability without a useful product, you don't really have much.

Q: It's rare for a major website to keep from bulking up over the years. How does Google's design stay simple?

We try to think long-term. When we place things on a page, it's not because we think we'll get an immediate payoff, but because that's the right place to put it. We will have new features, but I hope that those features are added so that they're more helpful than bulky. Like our language translation feature, where users can translate Google into their language. We wondered where should we put it on the website. Ultimately we put it in a small sentence on the Preferences page. It's just a small link that says "If you don't see your language here..." It's very small and unintrusive, but by cleverly placing it where people would most likely to want it, we've gotten 37,000 users to sign up to help us translate the site.

Our spellchecker is another significant piece of technology. It uses the "did you mean" link to suggest alternative spellings. It's so useful that the bulk it adds to the page is greatly outweighed by its usefulness.

As we add more features, eventually we'll have to evaluate all of them, to see if some aren't drawing enough usage, in which case we may take them away. We're placing new features carefully, and we're willing to pull some if they're not useful enough for our users.

Q: Can you say that you'd never go above, say, 20 links on the home page?

We can't say never. But I'll say that we'll be cautious. We have some idea of where our design can scale. Can we have a two-line footer? Maybe in the future. But generally we want to keep it sparse.

There's this one user, a Google zealot - we don't know who he is - who occasionally sends an e-mail to our "comments" address. Every time he writes, the e-mail contains only a two-digit number. It took us awhile to figure out what he was doing. Turns out he's counting the number of words on the home page. When the number goes up, like up to 52, it gets him irritated, and he e-mails us the new word count. As crazy as it sounds, his e-mails are helpful, because it has put an interesting discipline on the UI team, so as not to introduce too many links. It's like a scale that tells you that you've gained two pounds.

I hope that Google will still be here in 20 years. I don't know what the Google home page will look like that on that day. Maybe most of our users will be wearing computers, and will have a different interface to information and technology. Or there may be other mediums that we support. Like the Google news search [which just launched], news as a medium is different from search. We can't have a news home page that's so sparse that it doesn't have the news on it. We can't just have a search form, and ask users to guess the news.

Q: How do you handle user testing?

When I first started testing in 2000, we tested once a month. Now, we're user testing almost every week. We'll do a site-wide test once a month or so, with some tasks, but more free-form, just to see where people go, where they encounter problems. The other three weeks of the month, we test specific features. Adwords, for example, is a new product that's big enough that it needs its own test - it can't be layered into a sitewide test. So we test every 10 days, usually with eight users each. We want to find the big problems, and with eight users we definitely get to that level.

How do we get to the more granular problems? We get lots of data off the site, and from that we can see where traffic flow problems are. When we rolled out spellcheck, we had a link on the top saying, "If you didn't find what you're looking for..." A statistically significant number of users complained that they were still getting incorrect results. Turns out the site actually did tell them the correction, but they missed it because they went straight to the first result, which of course was wrong, and then they would click Back, scroll down to the bottom, and complain. So we thought, well, if we can't get them at the top, maybe we'll get them before they click the "complain" link at the bottom. So now we repeat the correction on the bottom of the page, and usage of the correction link has doubled. We hadn't realized we were missing half of the people who should have been clicking on that, until we watched the traffic. This is how we get to the subtleties on the site.

Q: Google runs no graphic ad banners. Can you describe your philosophy about online advertising? Is text more effective?

Our text-only ads outperform graphic ad banners. We feel no pressure at all to switch to graphic ads, either internally or from advertisers. It's a win-win. We want to present ads that are useful to our users, and advertisers can run things that are more targeted. They don't have to pay for production of the ad - with hiring a graphic artist and doing the graphics and animation, the creative cost can be very high. There's also the cost of development time. And ultimately you just get one ad out of it.

With text, all you need is a really good copy editor and you can test multiple versions. We have advertisers that run 10,000 versions of one ad. For example, Amazon is one advertiser. The ads can say "buy toasters at Amazon", or "buy clock radios at Amazon," and they didn't have to pay $2,000 to create each version of the ad. Our ad guidelines say that the ad should be targeted to this query, and the link should go to the deepest possible spot in the website. Like the toaster link should go to the toaster product page, not the Amazon home page.

Q: Then how do you guard against the irrelevant automated text ads I've seen on other sites, like "buy John Smith at Amazon"?

We don't let advertisers plug the search query into the ad. They have to write the ad beforehand. They can't run a "mad-libian" ad. Our theory is, if you need the user to tell you what you're selling, then you don't know what you're selling, and it's probably not going to be a good experience.

Q: What's next for Google?

A lot of exciting things. There are a couple of launches this fall that I'm excited about. Keep your eyes peeled. reports on the launch of Google News (Sept 23)

Back to Good Experience

Get Good Experience by e-mail:

- - - - - - - - - - - -

Get e-mail updates of e-mail

Copyright 1999-2003, Good Experience, Inc.